prevents OOM during quantizing LTX 2.3 combined model

[FurkanGozukara]

I had OOM on RTX 5090 and 96 GB VRAM

This fixed it

[silveroxides]

No this PR is poorly scoped, introduces arbitrary bloat via the filter aliases and lacks a full report from your OOM.

I have done learned rounding quantization on LTX 2.3 on 16GB VRAM and 96GB RAM without issues simply by using the far superior memory saving implementation already in place with the --low-memory flag. Peak VRAM load = 7.4GB

ctq -i ./ltx-2.3-22b-distilled.safetensors -o ./ltx-2.3-22b-distilled_int8tensorwisemixed_learned.safetensors --int8 --scaling_mode tensor --comfy_quant --save-quant-metadata --ltxv2 --optimizer prodigy --lr_schedule adaptive --lr_adaptive_mode "simple-reset" --lr_cooldown 0 --lr-shape-influence 1.75 --num_iter 9000 --calib_samples 4096 --lr 1.0 --top_p 0.5 --min_k 1024 --max_k 1024 --early-stop-stall 2000 --early-stop-lr 1.01e-8 --early-stop-loss 5e-9

Don't make PR containing arbitrary changes that don't take the full scope and complexity into consideration.